Êtes-vous un étudiant de l'EPFL à la recherche d'un projet de semestre?
Travaillez avec nous sur des projets en science des données et en visualisation, et déployez votre projet sous forme d'application sur Graph Search.
In this work we present a novel concept of augmenting a search tree in a packet-processing system with an dditional data structure, a Network of Shortcuts, in order to adapt the search to current input traffic patterns and significantly speed-up the frequently traversed search-tree paths. The method utilizes node statistics gathered from the tree and periodically adjusts the shortcut positions. After an overview of tree-search methods used in networking tasks such as lookup or classification, and a discussion of the impact of typical traffic characteristics, we argue that adding a small number of direct links, or shortcuts, to the few frequently traversed paths can significantly improve performance, at a very low cost. We present a shortcut-placement heuristic, compare our method to a standard caching mechanism and show how the use of different levels of aggregation in a search tree enables to achieve similar results with much fewer entries.
Amirkeivan Mohtashami, Dan Alistarh